MCP servers that manage Git & repos
Create issues & PRs, read commits, manage branches across GitHub/GitLab.
5 servers · Last updated June 17, 2026
TL;DR: These servers let an agent work with version control — opening issues and pull requests, reading commits and diffs, managing branches. They're the core of coding and DevRel agents. Official vendor servers (GitHub, GitLab) tend to expose the richest, best-maintained tool sets.
Bottom line: if you only try one, Git (Reference) is the most popular, verified option for this (74,000★). 4 more compared below.
Compare 5 servers
| Server | Transport | Auth | Verified | Stars | Tools for this |
|---|---|---|---|---|---|
| Git (Reference) | Local (stdio) | No auth | 74,000 | git_commit, git_create_branch, git_branch | |
| GitHub MCP Server | Local (stdio) | API key | 30,600 | issues (toolset), pull_requests (toolset) | |
| GitLab MCP Server (zereight) | Local (stdio) | API key | 1,700 | create_merge_request, get_merge_request_diffs, create_merge_request_note +4 | |
| Sentry MCP Server (remote) | Remote (HTTP) | OAuth | 1,100 | get_issue_details, search_issues, analyze_issue_with_seer +1 | |
| Docker Hub MCP Server | Local (stdio) | API key | 250 | get_repository_info, check_repository, check_repository_tag +5 |
The servers
Official MCP server for reading, searching, and manipulating a local Git repository's files and history.
GitHub's official server for repos, issues, PRs, and Actions — local Docker or hosted remote.
Popular community GitLab MCP server for projects, MRs, issues, pipelines, wikis and releases.
Official hosted Sentry MCP for investigating issues, errors, traces and performance data.
Official Docker Hub MCP server for searching images, reading metadata and managing repositories.
Use these in a stack
FAQ
Official GitHub MCP vs community ones?
GitHub's official server is actively maintained with toolset configuration and the broadest coverage; community servers can be lighter or target self-hosted setups.
Can these review code, not just read it?
Many expose PR review/comment tools in addition to read — combine with a code-search or filesystem server for deeper review agents.